Abstract
Foreign exchange windfalls such as those from natural resource revenues change nonresource exports, imports, and the capital account. The paper studies the balance between these responses and shows that the response to $1 of resource revenue is, for our preferred estimates, to decrease nonresource exports by 74 cents and increase imports by 23 cents, implying a negligible effect on foreign savings. The negative per $1 impact on exports is larger for manufactures than for other sectors, and particularly large for internationally mobile manufacturing sectors. Although standard Dutch disease analysis points to contraction of the tradable sector as a whole, division into nonresource exports and imports is important if, as suggested by much development literature, a higher share of exports to GDP is associated with faster growth. The large negative impact of resources on these exports points to the difficulty resource-rich economies face in diversifying their exports.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
We focus on nonrenewables, defined as fuels plus minerals, in line with the “resource curse” literature. Such exports (in particular their time variation) can reasonably be taken to be exogenous, as discussed in the section “Endogenous resource exports.”
An exception is Chen and Rogoff (2003).
There is now also a literature on subnational impacts (U.S. counties), prompted largely by the recent oil and gas boom in the United States. Jacobsen and Parker (forthcoming) and Michaels (2011) study the impact on a variety of outcome measures, while Allcott and Keniston (2014) study responses in manufacturing industries. They find that resource booms have a positive effect on local manufacturing, principally through the supply of inputs to the sector.
We are also able to back out the effects on the nonresource balance, NRB=X−M. The current account, S, is defined as: S=NRB+RNET+NY+NCT, where NRB is net nonresource exports, RNET is net resource exports, NY is net income from abroad and NCT is net current transfers from abroad (including workers’ remittances).
We choose this time period for defining net resource exporters because of the unbalanced character of our panel. See Table A2 for details. We provide robustness checks to other sample definitions in Appendix VI.
The model focuses on “the spending effect” and abstracts from the “resource movement effect” (Corden and Neary, 1982). The two effects are qualitatively the same for our objects of interest: contraction of the nonresource tradable sector and division of the contraction between imports and exports. Our econometric estimates capture the combined effect.
This is without loss of generality, merely reflecting the units in which labor and nontradables are measured.
We take this short-cut purely in order to simplify exposition of a well-known model.
In this case we add area, each countries endowment of land, as a control.
Pesaran and Smith (1995) show that for cross-sectional estimates based on time-averages of I(1)-variables, one does not need to worry about spurious correlation. The cross-sectional estimate is one way to get at the long-run estimate. Our panel estimates can be seen as the first step in the original Engle and Granger (1987) approach to cointegration and is an alternative route to the long-run estimate.
For a discussion of the exogeneity of resource exports, see van der Ploeg and Poelhekke (2010).
The p-value for the resource exports coefficient in the nonresource imports equation is 0.127. All tables in this paper use standard errors clustered at the country-level to account for potential serial correlation in the residuals. If we instead use robust standard errors, the resource exports coefficient becomes significant at the 1 percent level.
Hsiao, C. (1997) discusses identification under cointegration.
Although for oil, OPEC may be able to affect the price. The empirical evidence casts doubt on its ability to do so (see Barsky and Kilian, 2004 and Hamilton, 2008).
Note that the instruments are very strong when we use robust standard errors instead of standard errors clustered at the country-level, with F-statistics of 34 and 41 for the nonresource exports and imports relationships, respectively.
In a complementary exercise, valid under the assumption that the quantity of resource exports is exogenous, we allowed for separate effects of the quantity and price of resource exports (where the quantity was defined as the value of resource exports deflated by the resource price index). The elasticity w.r.t. quantity was estimated to be larger (more negative) than the elasticity w.r.t. to the price in the exports equation, −0.33 vs. −0.27, but the difference was not statistically significant. The unit changes were −83 and −66 cents per dollar resource exports. The quantity and price elasticities estimated in the imports equation were 0.08 and 0.14, translating into 23 cents and 44 cents per dollar resource exports. With robust standard errors, this difference was statistically significant at the 8 percent level and all four elasticities were significant at least at the 5 percent level. Kilian, Rebucci, and Spatafora (2009) investigate the effects of different types of oil price shocks on the external balances of oil exporters as a group and oil-importers as a group.
The optimal intertemporal responses for a country facing foreign exchange windfalls are similar to those from fiscal revenue windfalls from natural resource extraction. See Harding and van der Ploeg (2013) for theory and evidence on the latter. We leave it for future research to identify present value effects in our context.
Cross-equation tests confirm statistically significant difference between β Xaf and β Xma (χ2(1)=7.80, p=0.01), but not between β Xma and β Xsv (χ2(1)=1.52, p-value=0.22).
Using robust standard errors instead of standard errors clustered at the country-level, the following additional sectors have a significant negative coefficient: textile products, plastics, computers, and furniture. Chemicals was still the only sector with a significant positive coefficient. Estimation results available on request from the authors.
The variable we label TC is the constant distance elasticity, ηlog−log, estimated by Holmes and Stevens (2014). Following Allcott and Keniston (2014), we take 0.8 as threshold of “highly tradable goods.”
These results are consistent with Alcott and Keniston (2014), who look at the effect of resource booms across U.S. counties. Some of their effects may reflect relocation of economic activity within the United States, while our estimates are based on national data.
See Table A2 for a list of the countries.
See Ross (forthcoming) for a survey of this extensive literature.
Interaction variables are measured in deviations from their sample means, that is, the coefficient on resource exports is the effect at the mean of the interacted variable. All interaction variables are time-invariant, so the country fixed effects capture their direct effect and there is no need to include them separately in the regressions. We focus on “Rule of Law” and “Control of Corruption” from the Worldwide Governance Indicators (WGI).
The correlation coefficient between the rule of law and share of manufactures in nonresource export is 0.27 (−0.35 with share of food and agriculture, 0.06 with services).
Interaction coefficients for exports of the two other categories and for imports were not significant. The coefficient on ln R was always negative and significant for exports. For imports, it was always positive and often significant.
As we saw in the model of the section “The model,” a higher value of σ implies a higher price elasticity of export demand, and that more of the impact falls on a reduction in exports.
See Eberhardt and Teal (2011) for a discussion.
See: www.publications.worldbank.org/WDI/.
See: www.wits.worldbank.org/wits/ and Appendix IV for details.
For the concordance, see: www.nber.org/lipsey/sitc22naics97/.
See: www.unstats.un.org/unsd/snaama/dnlList.asp.
See: www.info.worldbank.org/governance/wgi/resources.htm.
References
Acemoglu, D., S. Johnson, and J. Robinson, 2005, “Institutions As the Fundamental Cause of Long-Run Economic Growth,” in Handbook of Economic Growth, ed. by P. Aghion, and S. Durlauf (Amsterdam, The Netherlands: Elsevier B.V).
Allcott, Hunt and Daniel Keniston, 2014, “Dutch Disease or Agglomeration? The Local Economic Effects of Natural Resource Booms in Modern America,” NBER Working Paper 20508.
Andersen, J. J. and S. Aslaksen, 2008, “Constitutions and the Resource Curse,” Journal of Development Economics, Vol. 87, No. 2, pp. 227–46.
Anderson, J. E. and E. van Wincoop, 2003, “Gravity and Gravitas: A Solution to the Border Puzzle,” American Economic Review, Vol. 93, No. 1, pp. 170–92.
Arezki, Rabah, Kirk Hamilton, and Kazim Kazimov, 2011, “Resource Windfalls, Macroeconomic Stability and Growth; The Role of Political Institutions,” IMF Working Paper 11/142.
Arezki, Rabah and Kareem Ismail, 2013, “Boom-Bust Cycle, Asymmetrical Fiscal Response and the Dutch Disease,” Journal of Development Economics, Vol. 101, No. C, pp. 256–67.
Arezki, Rabah, Valerie A. Ramey, and Liugang Sheng, 2015, “News Shocks in Open Economies: Evidence from Giant Oil Discoveries,” Working Paper UCSD.
Barsky, R.B. and L. Kilian, 2004, “Oil and the Macroeconomy Since the 1970s,” Journal of Economic Perspectives, Vol. 18, No. 4 115–134.
Bazzi, Samuel and Christopher Blattman, 2014, “Economic Shocks and Conflict: Evidence from Commodity Prices,” American Economic Journal: Macroeconomics, Vol. 6, No. 4, pp. 1–38.
Beine, M-, C. S. Bos, and S. Coulombe, 2012, “Does the Canadian Economy Suffer from Dutch Disease?,” Resource and Energy Economics, Vol. 34, No. 4, pp. 468–92.
Bernard, A., B. Jensen, S. Redding, and P. Schott, 2012, “The Empirics of Firm Heterogeneity and International Trade,” Annual Review of Economics, Vol. 4, No. 1, pp. 283–313.
Cashin, P., L. F. Cespedes, and R. Sahay, 2004, “Commodity Currencies and the Real Exchange Rate,” Journal of Development Economics, Vol. 75, No. 1, pp. 239–68.
Chen, Y. and K. Rogoff, 2003, “Commodity Currencies,” Journal of International Economics, Vol. 60, No. 1, pp. 133–60.
Corden, W.M. and J.P. Neary, 1982, “Booming Sector and De-Industrialisation in a Small Open Economy,” Economic Journal, Vol. 92, No. 368, pp. 825–48.
De Hoyos, R. E. and V. Sarafidis, 2006, “Testing for Cross-Sectional Dependence in Panel-Data Models,” The Stata Journal, Vol. 6, No. 4, pp. 482–96.
Disdier, A. and K. Head, 2008, “The Puzzling Persistence of the Distance Effect on Bilateral Trade,” The Review of Economics and Statistics, Vol. 90, No. 1, pp. 37–48.
Eberhardt, M. and F. Teal, 2011, “Econometrics for Grumblers: A New Look at the Literature on Cross-Country Growth Empirics,” Journal of Economic Surveys, Vol. 25, No. 1, pp. 109–55.
Ederington, Josh, Arik Levinson, and Jenny Minier, 2005, “Footloose and Pollution-Free,” The Review of Economics and Statistics, Vol. 87, No. 1, pp. 92–99.
Engle, R. F. and C. W. J. Granger, 1987, “Co-Integration and Error Correction: Representation, Estimation, and Testing,” Econometrica, Vol. 55, No. 2, pp. 251–76.
Feenstra, Robert C., 2004, Advanced International Trade: Theory and Evidence (Princeton and Oxford: Princeton University Press).
Hamilton, J. D., 2008, “Understanding Crude Oil Prices,” NBER working paper series, working paper 14492 (National Bureau of Economic Research).
Harding, T. and F. van der Ploeg, 2013, “Official Forecasts and Management of Oil Windfalls,” International Tax and Public Finance, Vol. 20, No. 5, pp. 827–66.
Hausmann, R., L. Pritchett, and D. Rodrik, 2005, “Growth Accelerations,” Journal of Economic Growth, Vol. 10, No. 4, pp. 303–29.
Hausmann, R., J. Hwang, and D. Rodrik, 2007, “What You Export Matters,” Journal of Economic Growth, Vol. 12, No. 1, pp. 1–25.
Holmes, Thomas J. and John J. Stevens, 2014, “An Alternative Theory of the Plant Size Distribution, with Geography and Intra- and International Trade,” Journal of Political Economy, Vol. 122, No. 2, pp. 369–421.
Hsiao, C., 1997, “Statistical Properties of the Two-Stage Least Squares Estimator Under Cointegration,” Review of Economic Studies, Vol. 64, No. 3, pp. 385–98.
Imbs, J., H. Mumtaz, M.O. Ravn, and H. Rey, 2005, “PPP Strikes Back: Aggregation and the Real Exchange Rate,” Quarterly Journal of Economics, Vol. 120, No. 1, pp. 1–43.
International Monetary Fund. 2007, Guide on Resource Revenue Transparency (Washington, DC: IMF).
International Monetary Fund. 2012, Macro-Economic Policy Frameworks for Resource Rich Developing Countries (Washington, DC: IMF).
Ismail, K. K., 2011, “The Structural Manifestation of the ‘Dutch Disease’ in Oil-Exporting Countries,” mimeo, International Monetary Fund.
Jacobsen, Grant D. and Dominic P. Parker, forthcoming, “The Economic Aftermath of Resource Booms: Evidence from Boomtowns in the American West,” The Economic Journal, http://onlinelibrary.wiley.com/doi/10.1111/ecoj.12173/abstract.
Johnson, S., J. Ostry, and A. Subramanian, 2006, “Levers for Growth,” Finance and Development, Vol. 43, No. 1, pp. 28–31.
Jones, B. F. and B. A. Olken, 2008, “The Anatomy of Start Stop Growth,” Review of Economics and Statistics, Vol. 90, No. 3, pp. 582–87.
Kilian, L., A. Rebucci, and N. Spatafora, 2009, “Oil Shocks and External Balances,” Journal of International Economics, Vol. 77, No. 2, pp. 181–94.
Mehlum, H., K. Moene, and R. Torvik, 2006, “Institutions and the Resource Curse,” The Economic Journal, Vol. 116, No. 508, pp. 1–20.
Michaels, Guy, 2011, “The Long Term Consequences of Resource-Based Specialisation,” The Economic Journal, Vol. 121, No. 551, pp. 31–57.
Pesaran, M. H., 2006, “Estimation and Inference in Large Heterogeneous Panels with a Multifactor Error Structure,” Econometrica, Vol. 74, No. 4, pp. 967–1012.
Pesaran, M. H. and R. Smith, 1995, “Estimating Long-Run Relationships from Dynamics Heterogeneous Panels,” Journal of Econometrics, Vol. 68, No. 1, pp. 79–113.
Prati, A. and T. Tressel, 2006, “Aid Volatility and Dutch Disease: Is There a Role for Macroeconomic Policies?” IMF Working Papers 06/145 (Washington, DC: International Monetary Fund).
Rajan, R. G. and A. Subramanian, 2011, “Aid, Dutch Disease, and Manufacturing Growth,” Journal of Development Economics, Vol. 94, No. 1, pp. 106–18.
Redding, S. and A.J. Venables, 2004, “Economic Geography and International Inequality,” Journal of International Economics, Vol. 62, No. 1, pp. 53–82.
Ross, M. forthcoming, “The Politics of the Resource Curse: A Review,” in Handbook on the Politics of Development, ed. by C. Lancaster, and N. van de Walle (Oxford University Press).
Sachs, J.D. and A.M. Warner, 1997, “Natural Resource Abundance and Economic Growth,” in Leading Issues in Economic Development, ed. by G. Meier, and J. Rauch (Oxford: Oxford University Press).
Santos Silva, J.M.C. and S. Tenreyro, 2006, “The Log of Gravity,” The Review of Economics and Statistics, Vol. 88, No. 4, pp. 641–58.
Stijns, J., 2003, “An Empirical Test of the Dutch Disease Hypothesis using a Gravity Model of Trade,” International Trade 0305001, EconWPA.
van der Ploeg, F., 2011, “Natural Resources: Curse or Blessing?,” Journal of Economic Literature, Vol. 49, No. 2, pp. 366–420.
van der Ploeg, F. and S. Poelhekke, 2010, “The Pungent Smell of ‘Red Herrings’: Subsoil Assets, Rents, Volatility and the Resource Curse,” Journal of Environmental Economics and Management, Vol. 60, No. 1, pp. 44–55.
van der Ploeg, F. and S. Poelhekke, 2013, “Do Natural Resources Attract Non-Resource FDI?,” The Review of Economics and Statistics, Vol. 95, No. 3, pp. 1047–65.
van der Ploeg, F. and A.J. Venables, 2013, “Absorbing a Windfall of Foreign Exchange; Dutch Disease Dynamics,” Journal of Development Economics, Vol. 103, pp. 229–43.
Additional information
*Torfinn Harding is an associate professor in the Department of Economics, Norwegian School of Economics (NHH). He is associated with Oxcarre/University of Oxford, University of Stavanger, and CESifo. He holds a PhD in Economics from the Norwegian University of Science and Technology. His research covers FDI promotion, export upgrading, and effects of natural resource exports. Anthony J. Venables is Professor of Economics at the University of Oxford where he also directs the Centre for the Analysis of Resource Rich Economies. He is a Fellow of the British Academy and of the Econometric Society. He has published extensively in the areas of international trade and spatial economics, including work on trade and imperfect competition, economic integration, multinational firms, and economic geography. This work was supported by the BP-funded Centre for the Analysis of Resource Rich Economies (Oxcarre), Department of Economics, University of Oxford. The authors would like to thank Silja Baller for research assistance and Roberto Bonfatti, Ghada Fayad, Keith Head, Rick van der Ploeg, L. Alan Winters, a referee and seminar participants at BI in Oslo, Global Area Conference CESifo, EEA in Oslo, ESEM in Malaga, ETSG in Leuven, NAE in Stavanger, NHH in Bergen, University of Oslo, Oxcarre/Dubai School of Government conference in Dubai, University of Oxford, RMET at UBC and Statistics Norway for useful comments.
An erratum to this article is available at http://dx.doi.org/10.1057/s41308-017-0034-4.
Electronic supplementary material
Appendices
Appendix I
A Three-Good Model
Distinct nonresource export and import response can be derived from a three-sector model with nontradables (price p n ), exportables (p x ), and import competing goods (p m ). Resource revenue is R, the economy’s expenditure function is e(p n , p x , p m )u where u is utility, and the revenue (or GNP) function is r(p n , p x , p m ); fixed endowments of factors are suppressed in the notation. Assuming for simplicity that there is no asset accumulation, the budget constraint is
Nontraded goods market clearing is
where subscripts denote partial derivatives. Prices of tradable goods, p x , p m , are fixed. These two equations implicitly define the two endogenous variables, p m and u, as a function of R. The effect of variation in R can be found by totally differentiating and rearranging to give (du)/(dR)=(1)/(e), (dp n )/(dR)=(1)/(r nn −e nn ).(e n )/(e)>0. Nonresource exports are X=r x −e x u and imports M=r m −e m u. Totally differentiating and using expressions for the change in p n and u,
The first expression on the right-hand side of each of these expressions is a relative price effect giving the general equilibrium effect of a change in the price of nontradables on supply and demand for the export and import competing good; in the first expression this is generally negative, and in the second positive. The second terms are income effects and, once again, for normal goods have negative on exports and positive on imports. It follows from homogeneity of revenue and expenditure functions that d(M–X)/dR=1.
Appendix II
Resources and Equilibrium
Equilibrium conditions (equations (8), (9), (10) and (11)) implicitly define the variables E1, n1, p1, and G1. We linearize around the equilibrium where R1=0, denoting the share of exports in tradable production by ς≡ X1/n1p1. Units are chosen such that L1=1, n1=μ, and hence E1=μp1 (equation (8)). Noting that R1=0 implies that X1=M1, equations (6) and (7) are ς=p1−σFMA1=G1σ−1FSA1. At this equilibrium p1, and G1 satisfy Equations (10) and (11) which both reduce to 1=μ(p1/G1)1−σ + ς. Evaluating differentials at these values of endogenous variables gives the following equations for the nonresource exports displaced by resource exports, −dX1/dR1, and the change in the real exchange rate, dp1/dR1:
An increase in R1 has a larger impact on X1 than on M1 if −dX1/dR1>1/2, that is, if ς[1+σ(1−2μ)]>1. Providing μ<1/2, this is satisfied for large enough ζ, σ.
Appendix III
Data
Data on GDP (GDP) and aggregate trade are from World Development Indicators (WDI).Footnote 29 Trade measures are exports and imports of goods and services (BoP), resource exports and imports covering fuel, metals, and ores (R is defined as gross resource exports); exports and imports of agriculture and food products (Xaf and Maf), and manufacturing products (Xma and Mma). Nonresource exports (X) are defined as: exports of goods and services (BoP) minus exports of fuel, metals, and ores; nonresource imports (M) are defined analogously. The nonresource balance (NRB) is an abbreviation for net nonresource exports (X-M). Exports and imports of services are defined as residuals: Xsv=X−Xaf−Xma; Msv=M−Maf-Mma.
Data on sector-specific exports are from Comtrade.Footnote 30 They cover the period 1984–2006 and are aggregated from SITC four-digit product classification to NAICS 1997 three-digit classification.Footnote 31
Data on bilateral nonresource exports used in the gravity estimation are also from Comtrade. The bilateral country-fixed variables used in the gravity estimations (distance and dummies for contiguity, common official primary language, and colonial relationship) and the unilateral country-fixed variables used in the regressions on aggregate data (area and dummies for landlocked and island status) are from CEPII.Footnote 32
The Comtrade data and the trade data from WDI are in our analysis measured in 2000 USD, that is, their original current USD values are deflated with the GDP deflator of the U.S. GDP, which is found by dividing U.S. GDP in current USD by U.S. GDP in fixed USD with year 2000 as the base year. Both GDP variables are from WDI. GDP is also measured in 2000 USD.
FMA (foreign market access) and FSA (foreign supplier access) are used as control variables. Appendix IV explains how they are constructed. They are denominated in 2000 USD.
Nonresource GDP, NRGDP, is calculated from National accounts data calculating value added and GDP from the production side, published by the UN. NRGDP is defined as total value added minus value added in Mining and Utilities (ISIC C and E).Footnote 33 The data are in 2005 USD.
The governance measures, Rule of Law and Control of Corruption, are from The Worldwide Governance Indicators (WGI) project, country averages over 1996–2006. Rule of law “captures perceptions of the extent to which agents have confidence in and abide by the rules of society, and in particular the quality of contract enforcement, property rights, the police, and the courts, as well as the likelihood of crime and violence.” Control of Corruption “captures perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as ‘capture’ of the state by elites and private interests.”Footnote 34
As a measure of transportation cost we use the log-log estimates presented by Holmes and Stevens (2014). We aggregate to NAICS 1997 three-digit and four-digit sectors by taking the simple average across all products within sectors.
See Tables A1 and A2 for descriptive statistics.
Appendix IV
Gravity Estimates
Analysis is based on the workhorse model of international trade flows, the gravity model. This states that bilateral exports between countries i and j, x ij , is a function of exporter country i characteristics, s i , importer country j characteristics, m j , and “between” country frictions
The focus of this paper is on countries’ exports and imports to and from all destinations, which we denote X i and M i , so
We proceed in two stages. First, we estimate the bilateral trade model in order to obtain values for the terms in the summation signs in Equation (A.2). Following the methodology of Redding and Venables (2004) this can be done using fixed effects for the country and importer characteristics, s i and m j , and the usual measures of proximity (distance, colony status, common language, contiguity) for the between-country frictions, t. We use nonresource trade and obtain estimates of foreign market access and foreign supplier access,
The former is a measure of how the fixed effects measuring foreign countries’ import demands, interacted with each countries’ proximity to country i, determine country i’s market access. The latter is analogous on the import side, measuring country i’s access to foreign sources of supply. Using these expressions, X i =s i FMA i , M i =m i FSA i .
We constructed annual bilateral nonresource trade flows by aggregating across all nonresource trade flows available at the SITC four-digit product level (also those smaller than 100,000 USD). We estimated a log-linear version of the gravity equation (A.1) on cross-sections covering all countries available except those with a population smaller than 0.5 million, starting with the first cross-section in 1970 and ending with the last in 2006. Hence we obtained 37 sets of coefficients. The dependent variable was log of exports from country i to j, ignoring zeros. As robustness, we did in early stages also experiment with the inclusion of zeros, estimating with the Pseudo Poisson Maximum Likelihood estimator (PPML) used by Santos Silva and Tenreyro (2006), but concluded that it would be unlikely to affect our results.
Table A3 presents statistics on the estimated coefficients of our components of t across the 37 cross-sections. Our estimates of the distance elasticity have a mean of −1.32. This agrees with the findings of Disdier and Head (2008), who found that the mean distance elasticity across 1,467 estimates in 103 papers was −0.9 with a standard deviation of 0.39, and that the distance elasticity has been relatively large since the middle of the twentieth century. The three bilateral dummy variables for colony, common language, and contiguity status show the expected positive sign.
Figure A1 presents the estimated coefficients and their standard deviations across the different cross-sections. The negative elasticity of distance has grown stronger over time. The elasticity of colony status decreased until the mid-1990s and has since been stable. The elasticities of common language and contiguity dummies fluctuate throughout the sample, but show now clear trend. It is important to notice that the gravity model is estimated on data from Comtrade, which only covers merchandise trade. However, we want to look at the impact of resource trade on all nonresource trade, services as well as merchandise. We assume that the measures FMA i and FSA i derived from merchandise trade are proxies for the impact of market access and supplier access on trade as a whole.
Estimated Gravity Coefficients and Corresponding std. Errors
Appendix V
Price Indices for Resource Trade
As an instrument for resource exports we use a price index based on country- and year-specific commodity price shocks constructed by Bazzi and Blattman (B&B, 2014). The shocks are calculated as (Bazzi and Blattman, 2014, p. 6): “the annual difference in each country’s log commodity export price index. Each country’s price index is a geometric average of all commodity export prices weighted by lagged export shares.” They use “U.S. dollar denominated prices from international markets.” We choose to use the version of their shock that gives the largest F-statistic in our first stage: the version leaves out 10 percent top potential price makers and makes no adjustments for the share of resource exports in GDP (this is as their “standard” measure, but the price index is not multiplied with the share of resource exports in GDP). We sum these shocks across time per country to get the price indices, which are the relevant measures for our long-run estimates. Note that our country fixed effects capture the level of the indices. When we use robust standard errors, the measure we chose has an F-statistic in the first stage of 34 and 41 in the exports and imports equations, respectively. The results with standard errors clustered at the country-level are presented in Table 3 in the main text.
Appendix VI
Further Econometric Issues
Panel-data unit root tests of the variables included in Equations (13) and (14) suggest in general that the series are integrated of order 1, that is, nonstationary in levels and stationary in first differences. Regarding the tests, we run the unit-root tests reported in lower part of Table A4. See the table-note for explanation. The tests are not always conclusive, but series are found to be integrated of the same order, that is, a unit root is often either rejected for all series by a particular test or not rejected for all series by a particular test. This is important as the key is that the series should be integrated by the same order for there to exist a stable relationship between them. As the panel-data unit root tests in the lower part of Table A4 show, we can reject the existence of a unit root in the residuals in Equations (13) and (14) in all tests. This indicates either a cointegrating relationship between nonstationary series integrated of the same order or a relationship between stationary variables. As is well known from the dynamic panel literature, the estimates can then be interpreted as the long-run relationship between the variables. We do the tests reported in Table A4 for all regressions using panel data and comment on the results in the note of each table throughout the paper. The full results of the cointegration tests are available upon request from the authors. We conclude that our estimates are not spurious due to the time-series properties of our variables. See van der Ploeg and Poelhekke (2013) for a recent application of these dynamic panel data procedures.
Tables A5, A6 and A7 are discussed in the section “Robustness” of the main text.
